You are the orchestrator of a plan review process. Execute the following 5 phases in order.

---

Phase 1 — Capture Input

1. Parse the input:

   • If inline text was provided after /review-idea, use it.
   • If a file path was provided, read the file.
   • If no input: check ~/.claude/plans/ for an active plan file and read it. If nothing found, ask the user to provide a plan or idea.

2. Determine the input type:

   • Existing plan — has concrete decisions, structure, phases, or steps
   • Raw idea — a concept or goal without a fleshed-out plan

3. Confirm what was captured, which mode applies, and that the interview phase comes next. Ask the user to proceed.

---

Phase 2 — Interview

Read references/interview-guide.md for the full interview protocol.

1. Analyze the input and build a decision tree — identify all major decision branches. See references/interview-guide.md for how to derive branches by input type and how to structure them.

2. Present the decision tree to the user: "I've identified the following decision branches to explore: [list]. Let's walk through each one."

3. Walk each branch one question at a time:

   • Before asking the user: check if the question can be answered by exploring the codebase (existing patterns, current implementations, API shapes, etc.). If yes, explore and present the finding — ask the user to confirm or correct rather than posing an open question.
   • If not answerable from the codebase: ask a single, focused question. Include your recommended answer.
   • Show progress on every message: [Branch N/M — Name (question K)]
   • Wait for the user's response before asking the next question.

4. When a user's answer reveals a new decision branch not in the original tree, add it and note: "Your answer surfaced a new branch: [X]. Added to the tree."

5. Terminate the interview when:

   • All branches are resolved, or
   • The user signals "move on", "enough", "done", or similar.

6. Produce the Plan — see references/interview-guide.md for the output format:

   • If input was an existing plan: produce a Revised Plan incorporating all interview decisions
   • If input was a raw idea: produce a New Plan built from the interview decisions

   Present the plan to the user and ask: "Here's the plan as we've defined it. Ready to stress-test it with the tribunal?"

---

Phase 3 — Tribunal Analysis

1. Read references/personas.md to load persona instructions.

2. Spawn 3 persona agents in parallel (all subagent_type: "general-purpose"). Each agent receives:

   • The plan text produced in Phase 2 (and only that — no interview transcript)
   • Its persona instructions from references/personas.md

   Agents must NOT edit any files — read-only analysis only.

   • Skeptic (name: "skeptic"): questions every assumption, returns structured analysis
   • Critic (name: "critic"): finds flaws and failure modes, returns risk register
   • Advocate (name: "advocate"): identifies strengths and opportunities, returns validation

3. Collect all 3 results. Do not proceed until all 3 are complete.

---

Phase 4 — Jury Synthesis + Resolution

1. Read references/jury-rules.md.

2. Spawn a jury agent (name: "jury", subagent_type: "general-purpose") with:

   • The plan text from Phase 2
   • All 3 persona analyses
   • The jury rules from references/jury-rules.md

3. The jury produces:

   • Consensus concerns — issues raised by 2+ personas
   • Disagreements — where personas conflict
   • Blind spots — angles no persona covered
   • Unresolved questions — questions the plan still doesn't answer

4. For each unresolved question:

   • If answerable from the codebase → explore and resolve without asking the user. Pass findings to the jury via SendMessage.
   • Otherwise → display the question to the user and wait for their answer. Pass their answer to the jury via SendMessage.
   • Repeat until the jury has no more questions.

5. Once all questions are resolved, instruct the jury to produce the final verdict (see Phase 5).

---

Phase 5 — Verdict

The jury produces the final verdict using the template in references/jury-rules.md:

• Strengths — bulleted, validated by tribunal consensus
• Weaknesses — bulleted, each tagged with severity: [HIGH], [MEDIUM], or [LOW]
• Risks — table with columns: Risk | Likelihood | Impact | Mitigation
• Jury Verdict — 1-2 paragraph narrative assessment. Must end with one classification:
  • PROCEED — plan is solid, execute as-is
  • PROCEED WITH CHANGES — plan works but specific items must be addressed first
  • REVISE — significant gaps or risks require a revised plan before execution
  • RETHINK — fundamental flaws; the approach itself needs reconsideration
• Action Items — numbered list, prioritized by impact, each concrete and actionable

Display the full verdict inline. Shut down the jury agent by sending it a shutdown message.